On Tue, Sep 7, 2010 at 6:03 PM, Dave Dyer <[email protected]> wrote: > > >I may agree, but otherwise I think you misunderstand completely > >what people are doing to improve their programs. > > A program developed by playing against itself, or any fixed reference > point, > can get stronger against that opponent while getting weaker against others. > > I'm floating the idea that a gaggle of MC based programs all playing each > other can collectively wander in an unproductive direction relative to human > players. >
You are of course talking about intransitivity. It does exist, but I do not understand why people make so much out of it. If your results against other programs goes up substantially, it's going to go up a LOT against humans. Maybe not quite as much, but a lot. All of the top computer chess developers use computer vs computer games to test and develop their programs. There is no other way, because it requires tens of thousands of games to measure small improvements. No human is going to sit down and let you play that many games against him. But my main point is that transitivity is pretty strong between computers and humans and what works against computers works against humans to a large extent. It is true that computer vs computer will distort the difference relative to humans but not by very much. But even if it were a lot, all that matters is that an improvement in one equals an improvement in the other and that is exactly how it works. It should be obvious to you that the new programs have achieved remarkable results against human players relative to the old style programs. > > _______________________________________________ > Computer-go mailing list > [email protected] > http://dvandva.org/cgi-bin/mailman/listinfo/computer-go >
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
